Topological Quantum Field Theories in Homotopy Theory...
Transcript of Topological Quantum Field Theories in Homotopy Theory...
Topological Quantum Field
Theories in Homotopy Theory
II
Ulrike Tillmann, Oxford
2016 V Congreso Latinoamericano de Matematicos
1
Leitmotif = Understanding Manifolds
1. Classical Cobordism Theory (Thom, ...)
2. Topological Field Theory (Witten, Atiyah, Segal, ...)
3. Cobordism Hypothesis (Baez-Dolan, Lurie, ...)
4. Classifying spaces of cobordism categories—— classifying space of cobordism categories. (Galatius–Madsen–Tillmann–Weiss)—— the cobordism hypothesis for invertible TQFTs—— filtration of the classical theory
5. Extracting information on Di↵(M)—— Mumford conjecture (Madsen–Weiss, ...)—— higher dimensional analogues. (Galatius–Randal-Williams, ...)
Manifolds
W is a manifold of dimension d if locally it is di↵eo-morphic to
Rd or Rd�0.
W is closed if it is compact and has no boundary.
Fundamental problem:– classify compact smooth manifolds W of dim d– understand their groups of di↵eomorphisms Di↵(W )– understand the topology of Mtop(W ) = BDi↵(W )– understand characteristic classes H⇤(Mtop(W ))
Topological moduli space
W closed of dimension d
Mtop(W ) := Emb(W,R1)/Di↵(W )
By Whitney, Emb(W,R1) ' ⇤ and hence
Mtop(W ) ' BDi↵(W )
• when W is oriented, the di↵eomorphisms preserve theorientation.• when @W 6= ;, the di↵eomorphisms fix a collar of theboundary.
Example 0: d = 0, W = n pts
Mtop(W ) = Emb(n pts, R1)/⌃n ' B⌃n
This is the configuration space of n unordered pointsin R1.
Example 1: d = 1, W = S1
Note that Di↵+(S1) ' S1 so that
Mtop(S1) ' BS1 ' CP1
Example 2: d = 2Compare with the moduli spaces of Riemann surfaces:
Mg = Tg/�g = R6g�6/�g
Note, for g > 1,
Di↵+(Fg)'�! ⇡0(Di↵+(Fg) =: �g
Hence, the forgetful map
Mtop(Fg) �!Mg
is a rational equivalence and
H⇤(Mtop(Fg);Q) = H⇤(Mg;Q)
Characteristic classes
There is a universal W -bundle on Mtop(W ), and anyfamily E of manifolds di↵eomorphic to W and parametrizedby a space X is given by a map
�E : X !Mtop(W )
Hence, the characteristic classes for W bundles aregiven by
H⇤(Mtop(W )) = H⇤(BDi↵(W ))
Construction of characteristic classes
Given a smooth orientable W -bundle
Wd ! E⇡! X
consider the vertical tangent bundle
Rd ! TvE :=a
x2XTEx ! E
It is classified by a map
�TvE : E ! BSO(d)
For every c 2 Hn(BSO(d)) consider
c(E) := ⇡!(�E(c)) 2 Hn�d(X)
Here ⇡! is the Gysin map. So
c 2 Hn�d(M(W )) = Hn�d(BDi↵(W ))
Example 2: d = 2
H⇤(BSO(2)) = H⇤(CP1) = Z[e]
For c = ei+1 2 H2i+2(CP1), c corresponds to the ithMorita-Mumford-Miller class
i := ei+1 2 H2i(Mg;Q)
Example 2: d = 2
H⇤(BSO(2)) = H⇤(CP1) = Z[e]
For c = ei+1 2 H2i+2(CP1), c corresponds to the ithMorita-Mumford-Miller class
i := ei+1 2 H2i(Mg;Q)
Homology stability [Harer, Ivanov, Boldsen, RW]For ⇤ < 2g/3, H⇤(B�g,n) is independent of g and n.
Mumford conjecture: For ⇤ (2g � 2)/3
H⇤(Mg;Q) ' Q[1,2, . . . ]
. early 1980s
4. Classifying space of cobordism categories
The enriched cobordism category Cobd:
Objects: R⇥`Md�1 M(M) ' `
Md�1 BDi↵(M)
Morphisms:
morCobd(M0,M1) '`Wd M(W ) ' `
Wd BDi↵(W ; @)
M0 �!W �M1
Example: d = 0
Objects: R as M�1 = ;
Morphisms:
morCobd(a0, a1) =a
nEmb(n pts, [a0, a1]⇥ R1)/⌃n
Classifying space functor
B : Topological Categories �! Spaces, C 7! BC
a a f b
a
f
b
g
c
gf
BC := (a
k�04k ⇥ {(f1, . . . , fk)|composable morphs})/ ⇠
For every object a in C, the adjoint of
41 ⇥morC(a, a) �! BCgives rise to a characteristic map
↵ : morC(a, a) �! maps([0,1], @;BC, a) = ⌦BC
Theorem [Galatius, Madsen, T., Weiss]
⌦B(Cob+d ) ' ⌦1MTSO(d) = limn!1⌦d+n((U?d,n)c)
where U?d,n is the orthogonal complement of the univer-
sal bundle Ud,n ! Gr+(d, n).
Note: Gr+(d, n) ,! Gr+(d, n+1) and U?d,n⇥R ,! U?d,n+1
Hence ⌃(U?d,n)c ! (U?d,n+1)c
Theorem [Galatius, Madsen, T., Weiss]
⌦B(Cob+d ) ' ⌦1MTSO(d) = limn!1⌦d+n((U?d,n)c)
where U?d,n is the orthogonal complement of the univer-
sal bundle Ud,n ! Gr+(d, n).
Note: Gr+(d, n) ,! Gr+(d, n+1) and U?d,n⇥R ,! U?d,n+1
Hence ⌃(U?d,n)c ! (U?d,n+1)c
Examples:
⌦B(Cob0) = ⌦1S1
Theorem [Galatius, Madsen, T., Weiss]
⌦B(Cob+d ) ' ⌦1MTSO(d) = limn!1⌦d+n((U?d,n)c)
where U?d,n is the orthogonal complement of the univer-
sal bundle Ud,n ! Gr+(d, n).
Note: Gr+(d, n) ,! Gr+(d, n+1) and U?d,n⇥R ,! U?d,n+1
Hence ⌃(U?d,n)c ! (U?d,n+1)c
Examples:
⌦B(Cob0) = ⌦1S1
⌦B(Cob+0 ) = ⌦1S1 ⇥⌦1S1
⌦B(Cob+1 ) = ⌦1+1S1
A similar theorem holds for any tangential structure ✓!
Excursion: tangential structures
Vectd(W ) = [W,BO(d)] = [W,Gr(d,1)]
E $ �E
Definition: Let ✓ : X ! BO(d) be a fiber bundle. A✓-structure on Wd is a lift of �TW : W ! BO(d) to X
Oriented +: Z/2Z! BSO(d)! BO(d)Framed fr: O(d)! EO(d)! BO(d)Background X: X ! BO(d)⇥X ! BO(d)Connected < k >: BO(d) < k >�! BO(d)
⌦1MT✓ = limn!1⌦d+n(✓⇤(U?d,n)c)
Example: ✓ : BO(d)⇥X ! BO(d)
⌦B(CobXd ) ' ⌦1MTSO(d)^X+ = limn!1⌦d+n((U?d,n)c^X+)
A ✓-structure on W is a map f : W ! X.
For d = 2, this gives a topological version ofGromov-Witten theory
Note: the cobordism category of manifolds in a back-ground space X gives rise to a generalised cohomology!
Lemma: H⇤(⌦10 MTSO(d),Q) ' ⇤⇤(H⇤>0(BSO(d);Q)[�d])
Lemma: H⇤(⌦10 MTSO(d),Q) ' ⇤⇤(H⇤>0(BSO(d);Q)[�d])
Proof: For fixed ⇤ > 0 and large n,
⇡⇤(⌦10 MTSO(d))⌦ Q = ⇡⇤( limn!1maps⇤(Sd+n, (U?d,n)c))⌦ Q
= ⇡⇤(maps⇤(Sd+n, (U?d,n)c)⌦ Q= ⇡⇤+d+n((Ud,n)
?)c ⌦ Q= H⇤+d+n((U
?d,n)
c)⌦ Q by Serre
= H⇤+d(Gr+(d, n))⌦ Q by Thom
= H⇤+d(BSO(d))⌦ QNow apply a theorem by Milnor-Moore on structure ofcommutative Hopf algebras and take duals.
Lemma: H⇤(⌦10 MTSO(d),Q) ' ⇤⇤(H⇤>0(BSO(d);Q)[�d])
For any closed d-dimensional manifold W
↵ : Mtop(W ) ⇢ morCobd(;, ;)! ⌦BCobd ' ⌦1MTSO(d)
For every c 2 H⇤+d(BSO(d))
↵⇤(c) = c 2 H⇤(Mtop(W );Q)
Example: For d = 2,
H⇤(BSO(2)) = H⇤(CP1) = Z[e], deg(e) = 2
So
H⇤(⌦10 MTSO(2),Q) = Q[1,2, . . . ], deg(i) = 2i
Fibration sequence
⌦1MTSO(d) !�! ⌦1⌃1(BSO(d)+) �! ⌦1MTSO(d�1).! is induced byU?d,n ! Ud,n ⇥Gr+(d,n) U
?d,n ' Gr+(d, n)⇥ Rd+n
Genauer proves that this corresponds to natural mapsof cobordism categories:
Cobd: d-dim cobordisms in [a0, a1]⇥ Rd+n�1 ⇥ (0,1)T
@ � Cobd: d-dim cobordisms in [a0, a1]⇥Rd+n�1⇥ [0,1)#Cobd�1: d� 1-dim cobordisms in [a0, a1]⇥Rd�1+n⇥ {0}
Application
Suppose Wd is an oriented manifold with boundary.Then restriction defines a map
r : BDi↵(W )! BDi↵(@W )
Let
c = pk11 . . . pkrr es 2 H⇤(BSO(d);Z) r = d/2 or (d+1)/2
Corollary [Giansiracusa-T.]
r⇤(c) = 0 when d even, or when d odd and s even.
In particular, 2i+1 goes to zero when restricted to thehandlebody subgroup Hg ⇢ �g.
Sketch of proofStarting with Madsen-Weiss, the proof has been con-tinuously simplified and the theorem generalised [Galatius–Madsen–T.–Weiss, Galatius, Galatius–Randal-Williams].
�d,n: space of all embedded d-manifolds without bound-ary which are closed as a subset in Rd+n; base point ?;topologized so that manifolds can disappear at infinity.
�kd,n: subspace of manifolds embedded in Rk⇥(0,1)n+d�k.
�d,n = �d+nd,n � · · · � �k
d,n � · · · � �0d,n '
a
W, @=?BDi↵(W )
Cobkd,n: k-fold cobordisms category of d-manifolds em-
bedded in Rd+n
limn!1 Cob1d,n = Cobd and limn!1 Cobdd,n = exCobd.
Step 1: (U?d,n)c ' �d+nd,n
tangential information: (P, v) 7! v � P .
Step 2: �kd,n
'�! ⌦�k+1d,n for k > 0
adjoint of R⇥�kd,n ! �k+1
d,n , (t,W ) 7!W+(0, . . . , t, . . . ,0)
Step 3: B(Cobkd,n) ' �kd,n
�1d,n ' B(Cob1d,n)
. �1d,n
' � Nq(P) '�! Nq(Cobd,n)
The fiber of the left arrow at W is the poset of regularvalues of the projection onto R; its nerve is contractible.
Step 1: (U?d,n)c ' �d+nd,n
tangential information: (P, v) 7! v � P .
Step 2: �kd,n
'�! ⌦�k+1d,n for k > 0
adjoint of R⇥�kd,n ! �k+1
d,n , (t,W ) 7!W+(0, . . . , t, . . . ,0)
Step 3: B(Cobkd,n) ' �kd,n
=) B(Cobkd,n) ' �kd,n ' · · · ' ⌦d+n�k�d+n
d,n ' ⌦d+n�k(U?d,n)c
for k = 1 and n!1, B(Cobd) ' ⌦1�1MTSO(d)for k = d+ n and n!1, B(exCobd) ' ⌦1�dMTSO(d)
An even finer filtration
⌦1MSO ' limn!1 lim
d!1⌦n(Un,d)
c
' limd!1
limn!1⌦n(U?d,n)c
' limd!1
limn!1B(Cobdd,n)
A 2-morphism in Cob22,1.
5. Unstable information
Question: Given Wd with ✓-structure, how close to anisomorphism in homology is ↵?
↵ : Mtop(W ) = BDi↵(W ) �! ⌦1MT✓(d)
5. Unstable information
Question: Given Wd with ✓-structure, how close to anisomorphism in homology is ↵?
↵ : Mtop(W ) = BDi↵(W ) �! ⌦10 MT✓(d)
Theorem [Barrett-Priddy, Quillen, Segal]
For d = 0: B⌃n↵�! ⌦1MTO(0) ' ⌦1S1 is a homol-
ogy isomorphism in degrees ⇤ n/2.
Theorem [Galatius, Madsen, T., Weiss]
⌦B(Cob@d) ' ⌦B(Cobd) ' ⌦1MTSO(d)
Cob@d contains only cobordisms W with⇡0(W )! ⇡0(M1) surjective
Theorem [Galatius, Madsen, T., Weiss]
⌦B(Cob@d) ' ⌦B(Cobd) ' ⌦1MTSO(d)
Cob@d contains only cobordisms W with⇡0(W )! ⇡0(M1) surjective
Previous result:
Theorem [T.] ⌦B(Cob@2) ' Z⇥B�+1
By homology stability, for ⇤ (2g � 2)/3
H⇤(B�+1) = H⇤(B�1) = H⇤(B�g)
Together these theorems imply
Theorem [Madsen-Weiss]
For d = 2: BDi↵(Fg)↵�! ⌦10 MTSO(2) is a homology
isomorphism in degrees ⇤ (2g � 2)/3.
=) Mumford’s Conjecture:
H⇤(Mg;Q)) ' H⇤(BDi↵(Fg),Q) ⇠ Q[1,2, . . . ]
Let Wg = ]gSd ⇥ Sd and d > 2. As Wg is (d � 1)-connected, it has a ✓ structure for
✓ : BO(2d) < d > �! BO(d)
Theorem [Galatius, Randal-Williams]For ⇤ < (g � 4)/2,
H⇤(BDi↵(Wg;D2d)) ' H⇤(BDi↵(Wg+1;D2d)
Theorem [Galatius, Randal-Williams]↵ : hocolimg!1BDi↵(Wg;D2d)! ⌦10 MT✓(d)is a homology equivalence
Similar results hold for more general simply connectedmanifolds of even dimension.
There are problems in odd dimensions!
Note
For d = 1: BDi↵(S1) ' CP1 ↵�! ⌦10 MTSO(1) '⌦1+1S1 is trivial in rational homology fro any W3.
There are problems in odd dimensions!
Note
For d = 1: BDi↵(S1) ' CP1 ↵�! ⌦10 MTSO(1) '⌦1+1S1 is trivial in rational homology fro any W3.
Theorem [Ebert]
For d = 3: BDi↵(W3) ↵�! ⌦1MTSO(3) is trivial inrational homology.
More generally, there are rational cohomology classesin the cohomology of ⌦1MTSO(2n + 1) that are notdetected by any 2n+1-dimensional manifold.
Qg = ]gSd ⇥ Sd+1
Theorem Perlmutter
For ⇤ (g � 3)/2,
H⇤(BDi↵(Qg;D2d+1) ' H⇤(BDi↵(Qg+1;D2d�1)
Hebestreit-Perlmutterconstruct a variant of the cobordism category that hasa good chance to play the role of Cobd in the odd di-mensional case.
Also true for other product of spheres.
General scheme of proof
1. Homology stability with respect to connected sumwith a suitable basic manifolds, e.g. Sd ⇥ Sd
2. Group completion argument to identify the ho-mology of the limit with the classifying space of asubcatgory of an appropriate cobordism category, e.g.limg!1H⇤BDi↵(Fg,D2) ' H⇤⌦BCob@2
3. Parametrised surgery to show that the subcate-gory has a classifying space that is homotopic to thewhole category, e.g. BCob@d ' BCobd
4. Identify the homotopy type of the classifying spaceof the whole category, e.g. ⌦BCobd ' ⌦1MTSO(d)
Multiplicative structure
⌃n ⇥⌃m �! ⌃n+m
Di↵(Fg,1)⇥Di↵(Fh,1) �! Di↵(Fg+h,1)
1 2 g
1 2 h
...
...
AutFn ⇥AutFm �! AutFn+m
. AutFn ' HtEq(VnS1)
Products
Gn ⇥Gm �! Gn+m
induce a monoid structure on
M :=G
n�0BGn
If products are commutative upto conjugation by anelement in Gn+m,
then the induced product on H⇤(M) is commutative
Group completion
Algebraic: M �! G(M) = Grothendieck group of MExample: N �! G(N) = Z
Homotopy theoretic:M �! ⌦BM = map⇤(S1, BM) = loop space of BM
• M = G a group =) ⌦BG ' G• M discrete =) ⌦BM ' G(M)
Group Completion Theorem:Let M =
Fn�0Mn be a topological monoid such that
the multiplication on H⇤(M) is commutative. Then
H⇤(⌦BM) = Z⇥ limn!1H⇤(Mn) = Z⇥H⇤(M1)
Wn H⇤-stab. ⌦B(tnM(Wn))
n pts n/2 ⌦1S1 Barratt-Eccles-Priddy
Sg,1 2g/3 ⌦1MTSO(2) Madsen-Weiss
Ng,1 g/3 ⌦1MTO(2) Madsen-Weiss & Wahl
VnS1 n/2 ⌦1S1 Galatius
#g(Sd ⇥ Sd)1 (g-4)/2 ⌦1MTSO(2d)hdi Galatius-Randal-Williams
(Sd ⇥Dd+1)1 (g-4)/2 Q(BO(2d+1)hdi) Botvinnik-Perlmutter
discrete (g-4)/2 ⌦1X�� Nariman
(Sd ⇥ Sd+1)1 (g-3)/2 candidate Perlmutter